Opinionated - China Chipping Away to Semiconductor Dominance

[JSCh]

China new fabs to join competition for foundry orders in late 2018
Monica Chen, Taipei; Willis Ke, DIGITIMES
Tuesday 14 August 2018

As most of the 12-inch wafer foundry fabs built in China in 2016-2017 will kick off commercial run in late 2018, China's combined monthly foundry capacities at 12-inch fabs will surge over 40% to near 700,000 pieces in 2018 from 2017, posing new competition pressure to non-China foundry houses including United Microelectronics, Vanguard Semiconductor International, Globalfoundries and Taiwan Semiconductor Manufacturing Company (TSMC), according to industry sources.

However, it remains to be seen as to whether the new 12-inch and 8-inch foundry fabs in China can effectively win patronage from chipmakers and secure viable operations, the sources said. While poor customer relationship will lead to low capacity utilization, such issues as how to improve yield rates and upgrade foundry technologies, high equipment depreciation cost and difficulty in recuriting high-end engineers will also constitute new variables to affect the survival of the new fabs.

In addition, the lingering silicon wafer supply shortfalls and ever-rising quotes will further affect the performance of the new foundry fabs in China. Taiwan's GlobalWafers, the world's number-three wafer maker, has seen its production capacities for 6-, 8-, and 12-inch wafers fully booked through the end of 2020, with new supply contracts for 2021-2025 showing no leeway for downward price adjustment. Japan's Sumco has recently noted that its silicon wafer quotes will pick up 20% on year in 2018 and rise further in 2029, and the company has also started to sign long-term contracts for wafer supply starting 2021, the sources indicated.

Conservative about third-quarter prospects

Foundry houses in Taiwan and China are conservative about performance prospects for the third quarter of 2018 amid market uncertainties including smartphone shipments little likely to grow in the high season and market scale of new chips remaining unclear.

TSMC, for instance, recently adjusted downward its annual revenue projections for 2018, due mainly to demand for crypto mining ASICs and GPU chips weakening sharply on plunges of the exchange prices for virtual currencies. UMC also expected flat shipment performance for the third quarter, while second-tier foundry Semiconductor Manufacturing International Corp (SMIC) in China estimated its revenues and gross margins to decline sequentially in the third quarter after posting a profit growth of over 40% in the second quarter.

Strong demand for AI-based high performance computing (HPC) chips and chips for automotive electronics, 5G and IoT applications is expected to give new growth momentum for foundry houses, industry sources said, adding that, however, smartphone chips now remain the largest revenue source for foundry firms despite a foreseeable slowdown in sales momentum for smartphones in the third quarter.

Orders for AI chips and new application chips have yet to be as robust as expected. Accordingly, it remains to be observed whether demand for diverse new chips can support effective capacities of advanced processes such as 7nm at TSMC and Samsung Electronics, the sources noted.

 

[JSCh]

Posted: Aug 09, 2018
Self-assembling single-digit nanometer memory cells

(Nanowerk News) An international collaborative research team has developed sub-10-nanometer-range spin transfer torque (STT) magnetic tunneling junctions (MTJs).

STT-MRAM (spin-transfer torque-magnetoresistive random access memory) has been intensively researched in recent years and recently been commercialized. The STT-MRAM is capable of replacing existing semiconductor-based working memories due to its excellent capabilities in terms of operation speed and read/write endurance.

Conventional STT-MRAM is based on out-of-plane magnetized MTJs in which the storage layer magnetization is pulled out-of-plane thanks to a perpendicular anisotropy. Design and fabrication of such a device with multi-level signal processing is a major milestone toward future energy-efficient memory and logic operations.

To increase the performance and capacity of STT-MRAM, it has been essential to make the MTJ smaller, while maintaining the capabilities to retain information and be switched by a small current. More recently, several research groups have reported on how to decrease the size of MTJs with good thermal stability by using shape anisotropy of the magnet.

UC Berkeley and Chinese researchers have demonstrated a memory cell to be controllable down to a sub-10-nanometer scale. The STT-MTJ is capable of replacing existing semiconductor-based devices due to its excellent capabilities in terms of power consumption, operating speed, and endurance. Moreover, it is nonvolatile, i.e., no power supply is required to retain stored information, making it indispensable for future ultralow-power electronics.

To further increase the performance and capacity of single-digit-nanometer-range MTJs, it is essential to have a good thermal stability factor of above 65, while maintaining the capabilities to retain information and be switched by very low energy. The demonstrated point contact MTJs shows a thermal stability factor of more than 80.
These research findings, published in the journal Applied Physics Letters ("Self-assembled single-digit nanometer memory cells"), are a major step toward building ultra-low-power STT-MRAM.

"This present work opens a path towards the fabrication of highly energy efficient spin memory devices," concludes Professor Jeongmin Hong from the School of Optical and Electronic Information, Huazhong University of Science and Technology. "Ultimately, we will be able to develop a spin computer the size of a fingernail to replace today's big supercomputer."

Source: Huazhong University of Science and Technology


https://www.nanowerk.com/nanotechnology-news2/newsid=50863.php

 

[JSCh]

Gree Sets Up Semiconductor Unit to Help China Make Chips at Home
LIAO SHUMIN
DATE: WED, 08/22/2018 - 15:03 / SOURCE:YICAI

Gree Sets Up Semiconductor Unit to Help China Make Chips at Home​

(Yicai Global) Aug. 22 -- Chinese appliance maker Gree Electric, best known for its air conditioners, has set up a CNY1 billion (USD146 million) chipmaking subsidiary after several claims that it would enter the semiconductor business as the country looks to reduce reliance on imports.

Gree is the sole shareholder in Zhuhai Zero Border Integrated Circuit, while the new unit’s legal representative is Dong Mingzhu -- Gree’s chair and one of China’s most successful businesswomen -- according to data from the National Enterprise Credit Information Publicity System. The firm’s registered business scope covers a broad range of areas, including chipmaking, the Internet of Things and software for mobile devices.

The firm announced in April that it would not pay out dividends for 2017, the first time in over a decade. Instead, Dong told shareholders that the company would be using the profit to foray into other sectors, particularly chipmaking, in which Dong said she was prepared to invest as much as CNY50 billion -- or more than double last year’s net income.

Semiconductors have been a hot topic in the world’s second-largest economy after the United States reprimanded one of China’s biggest telecoms equipment makers, ZTE, in April for illegally selling American technology to Iran and North Korea and failing to punish employees behind the scheme.

The US forbade domestic companies from selling technology to the firm, bringing ZTE to a standstill until American lawmakers lifted the ban after US President Donald Trump stepped in. ZTE may have escaped the ordeal, but the debacle highlighted major holes in China’s domestic chip development.

In order to nurture its own semiconductor unit, Gree plans to quadruple revenue to CNY600 billion (USD87.7 billion) within five years, Dong said in June, adding that by 2019 she expects all of the company’s air conditioners to be using its own chips. The half-decade goal might seem farfetched, but income soared 45 percent annually to CNY150 billion in 2017, with air-con sales making up about 82 percent of all revenue. If it wants to hit the target, it will need income to increase by an average of just under 32 percent a year.

 

[qwerrty]

People's Daily,China
‏Verified account @PDChina

Li-Fi in China: A Chinese homegrown, commercial-grade ultra-wideband visible light communication chipset was released in Chongqing on Friday. First of its kind in the world, it supports data transmission at speeds 10 times faster than 5G and will be put to trial use in 2 years.

https://twitter.com/x/status/1033250017698402304

---------------------------
赵伟国：紫光2019年量产64层128G存储器
http://www.52rd.com/S_TXT/2018_8/TXT108207.HTM?WebShieldDRSessionVerify=MMwgufUnrSuEiRQ70lMF
mass production of 32-layer nand at the end of this year
mass production of 64 layer nand in 2019
128 layers in development. skip 96-layer nand

 

[cirr]

People's Daily,China
‏Verified account @PDChina

Li-Fi in China: A Chinese homegrown, commercial-grade ultra-wideband visible light communication chipset was released in Chongqing on Friday. First of its kind in the world, it supports data transmission at speeds 10 times faster than 5G and will be put to trial use in 2 years.

https://twitter.com/x/status/1033250017698402304

---------------------------
赵伟国：紫光2019年量产64层128G存储器
http://www.52rd.com/S_TXT/2018_8/TXT108207.HTM?WebShieldDRSessionVerify=MMwgufUnrSuEiRQ70lMF
mass production of 32-layer nand at the end of this year
mass production of 64 layer nand in 2019
128 layers in development. skip 96-layer nand

Chongqing firm showcases chip 10 times faster than 5G

2018-08-27 11:21:18 Global Times Editor : Li Yan

China has launched the world's first commercial grade visible light communication chip that supports transmission as fast as 10 gigabytes a second, or 10 times a 5G mobile network.

The chip released Friday at the August 23-25 Smart China Expo, in Southwest China's Chongqing Municipality, uses a light-emitting diode to surf the internet.

The chip is "safe, stable, high-speed and efficient with a low cost," Wu Jiangxing, an academician of the Chinese Academy of Engineering, was quoted as saying by news site people.cn on Friday.

Visible light communication (VLC) is a telecommunication technology with low carbon emissions which could realize zero energy consumption, the report said.

It could also efficiently avoid electromagnetic signal leakage, a foible of conventional radio communication.

Mass production would bring a breakthrough to the telecommunication industry and set a milestone for building a smart home, Wu said.

Fu Liang, a Beijing-based telecoms industry expert, said that the technology is still in its experimental stage with problems to be resolved before mass application.

"Internet service providers and users must upgrade their devices to be compatible with VLC equipment," Fu told the Global Times. "It's not an easy or short-term process to complete the change."

Construction of a network and maintaining its high-speed bandwidth for households "would be another problem," Fu said.

The technology might be used in some industries, but it was too early to talk about its future in daily domestic life, Fu said.

http://www.ecns.cn/news/sci-tech/2018-08-27/detail-ifyxikfc9643442.shtml

 

[cirr]

性能最强的国产通用图形处理器JARI G12来啦！

中船重工第七一六研究所 7月27日

7月19日—20日，“2018自主可控计算机大会”在北京国际会议中心隆重召开，七一六所拥有自主知识产权的通用高性能图形处理器JARI G12闪耀全场，以其完全自主可控的设计、超强的功能性能、巨大的应用前景引起了现场的广泛关注。

JARI G12 GPU

七一六所自主研发的JARI G12是目前性能最强的国产通用图形处理器。该处理器采用混合渲染架构，兼顾数据带宽和渲染延时需求，极大的增强了芯片的灵活性和适应性；提供PCIe 3.0总线，支持x86处理器和龙芯、飞腾、申威等国产处理器；支持4路数字通道和1路VGA输出，提供DP、eDP、HDMI、DVI等通用显示接口，单路数字通道最大输出分辨率为3840×2160@60fps，支持扩展、复制显示和“扩展+复制”显示模式；内建视频编解码硬核，支持2路3840×2160分辨率视频的编码、解码功能；支持OpenGL 4.5和OpenGL ES 3.0，满足高性能3D加速和VR显示需求；支持OpenCL 2.0，满足并行计算和云计算的使用需求；集成张量加速计算硬核，支持AI计算加速。该GPU支持Windows、Linux、VxWorks等主流操作系统，同时支持中标麒麟、JARI-Works、道等国内自主可控操作系统，具备健全的生态环境体系。

目前，七一六所JARI G12通用图形处理器领先国内同行业一代，在功能性能上实现全面超越，引爆了整个会场，受到广泛关注，并收到中电集团、中航集团、国防科技大学等多家单位的试用和需求意向。

https://mp.weixin.qq.com/s?__biz=MzAwNzM2NjQxMQ==&idx=5&mid=2455506663&sn=8df35b3a46fc848a8bb87aa1a6438d04

 

[qwerrty]

100G silicon photonics chip for optical transceiver put into production developed by accelink and government research institutes.


-------------------------------------------------------------------------
中国自主研发的首款商用100G硅光芯片正式投产
来源: 人民网-人民日报 作者: 范昊天 2018-08-30 14:04

[导读]记者从中国信息通信科技集团获悉：日前，我国自主研发的首款商用“100G硅光收发芯片”正式投产使用。

原标题：我国首款商用100G硅光芯片投产

本报武汉8月29日电 （记者范昊天）记者从中国信息通信科技集团获悉：日前，我国自主研发的首款商用“100G硅光收发芯片”正式投产使用。该系列产品支持100—200Gb/s高速光信号传输，具备超小型、高性能、低成本、通用化等优点，可广泛应用于传输网和数据中心光传输设备。

据介绍，该款商用化硅光芯片由国家信息光电子创新中心、光迅科技公司等单位联合研制。在一个不到30平方毫米的硅芯片上，集成了包括光发送、调制、接收等近60个有源和无源光元件，是目前世界上集成度最高的商用硅光子集成芯片之一。

国家信息光电子创新中心专家委员会主任余少华说，此次100G硅光芯片的产业化商用，是硅光技术领域的新突破，表明我国已经具备了硅光产品商用化设计的条件和基础。借助集成电路已大规模商用的CMOS工艺平台实现硅光芯片的生产制造，可以有效解决我国高端光电子芯片制造能力薄弱、工艺能力不足的问题
http://finance.zijing.org/2018/0830/763436.shtml

------------------
huawei's caliopa

http://www.caliopa.com/technology

 

[Adam WANG SHANGHAI MEGA]

7nm制程A76架构 华为麒麟980正式亮相
2018-09-01 07:46
8月31日晚，德国柏林的IFA 2018大展上，余承东正式揭晓了华为的新一代旗舰级移动SoC处理器“麒麟980”， 新一代顶级人工智能手机芯片，拥有六项世界第一，性能更强，能效更高，设计更紧凑。

麒麟980是全球第一颗采用7nm新工艺的手机SoC，与台积电合作，2015年就开始相关研究，2016年构建IP单元，2017年进行工程验证，2018年投入量产，历时长达36个月。

为了研发麒麟980，华为和台积电投入了1000多名半导体设计和工艺高级专家 ，前后使用了5000多块工程验证板。

按照台积电的数据，7nm相比于10nm可带来20%的性能提升，40%的能效提升，60%的晶体管密度提升 ，而应用到麒麟980身上，集成了多达69亿个晶体管 ，相比麒麟970(10nm 55亿个)增加了25%，同时晶体管密度增加了55% ，而芯片面积仅相当于指甲盖大小。

CPU部分，麒麟980首发采用ARM Cortex-A76架构 ，比上代麒麟970 A73性能提升75%，能效提升58%。

麒麟980还设计了灵活的2+2+4核心配置、智能调度机制，包括两个2.6GHz A76大核心、两个1.92GHz A76中核心、四个1.8GHz A55小核心。

华为表示，相对于传统的大小核两档位设计，大中小核的能效架构提供了更为精确的调度层次，让CPU在重载、中载、轻载场景下灵活适配，无论是只发短信，还是边打手游边听歌，麒麟980都能精确调度，大大降低了CPU的实际功耗。

从华为给出的调度设计看，两个超大核心只会在图库、大型游戏、APP启动等高负载场景中才会启动。

相比于骁龙845，麒麟980号称性能高出37%，能效高出32%。

GPU方面，麒麟980首发商用ARM Mali-G76，十核心配置 ，相比麒麟970 Mali-G72MP12性能提升46%，能效提升178%，GPU Turbo加持后远超骁龙845。

麒麟980还创新性地引入了AI调频调度技术 ，能够实时学习帧率、流畅度和触屏输入变化，预测手机任务负载，动态智能感知手机使用过程中存在的性能瓶颈，及时进行调频调度。

在游戏场景下，使用AI调频调度技术预测游戏每帧负载，预测准确性可以提升30%以上，有效提升游戏平均帧率，大幅降低游戏抖动率，减少游戏卡顿。

拍照方面，麒麟980全新升级第四代自研ISP，而且首次配备双ISP，像素吞吐率提升46% ，能够分区域调节图像色彩与灰阶，支持更多摄像头(暗示Mate 20后置三摄？)，同时还有全新HDR色彩还原，录像的时候能效提升23%，延迟降低33%。

AI方面，首次集成双NPU单元，强大算力支持人脸识别、物体识别、物体检测、图像分割、智能翻译等AI场景，并有更高精度的深度网络，具备更佳的实时性。

无论是表演节奏感极强的舞蹈，还是在镜头前快速跑步，麒麟980都能实时绘制出人体的关节和线条，而物体检测能力可以准确识别多种物体，拍照预览的实时物体跟踪。

官方宣称，麒麟980每分钟可识别图像4500张，速度提升120%，而且是骁龙845的几乎两倍。

网络通信方面，麒麟980率先支持LTE Cat.21，业界最高下行速率1.4Gbps，包括4x4 MIMO 1.2Gbps(三载波聚合)、2x2 MIMO 200Mbps两部分组成，并支持256QAM，是目前最先进的4.5G技术。

同时配套使用全球最快的手机Wi-Fi无线芯片Hi1103 ，率先支持160M带宽，理论峰值下载速率可达1.7Gbps，是业界同期水平的1.7倍。

Hi1103还支持L1+L5双频GPS精准定位 ，L5频段定位精度提升10倍，在地形复杂的城市峡谷、高架道路等地区也可精准导航。



此外，麒麟980还率先支持LPDDR4X-2133高频内存，整合i8传感处理器，支持UFS 2.1存储规格。

麒麟980将在Mate 20系列上首发，10月16日在英国伦敦正式发布。

 

[JSCh]

Huawei promises its 7nm Kirin 980 processor will destroy the Snapdragon 845
More power, delivered more efficiently
By Vlad Savov@vladsavov Aug 31, 2018, 8:30am EDT

The Kirin 980 system-on-a-chip.
Image: Huawei
At IFA today, Huawei announced its newest system-on-a-chip, the Kirin 980, which boasts a number of world firsts. It’s the first 7nm mobile processor, the first one built around ARM’s Cortex-A76 CPU and Mali-G76 GPU, the first with a Cat.21 smartphone modem supporting speeds up to 1.4Gbps, and the first chip to support 2,133MHz LPDDR4X RAM. The Kirin 980 has 6.9 billion transistors, but I’ve seen it for myself and it’s no larger than a thumbnail.

The road to today’s announcement started three years ago for Huawei, with the company engaging more than 1,000 senior semiconductor design experts and churning through more than 5,000 engineering prototypes. The end result is roughly a 20 percent speed improvement and a 40 percent reduction in power consumption relative to Huawei’s previous generation.

QUALCOMM HAS BEEN PUT ON NOTICE
But the product that Huawei really wants to compare the Kirin 980 against is Qualcomm’s Snapdragon 845, the chip that figures in practically every Android flagship phone not made by Huawei. It’s worth noting that the 845 has been out for many months now whereas Huawei’s 980 won’t be in any retail devices until next month at the earliest (Huawei let slip that it’s planning its Mate 20 announcement for October 16th). Still, the margin of improvement that Huawei is quoting over its major rival is impressive.

On the memory front, Huawei says the Kirin 980 has 20 percent better bandwidth and 22 percent lower latency than the Snapdragon 845. In practical terms, that means faster app launches across the full range of the world’s most popular apps. In gaming applications, the 980 has been shown to produce 22 percent higher frame rates than the 845, and its power consumption when gaming is said to be 32 percent lower.

Photography performance is another major upgrade for the Kirin chip, according to Huawei’s numbers. Using a new dual ISP (image signal processor), the Kirin 980 is 46 percent faster at camera processing than its predecessor, with a related 23 percent improvement in power efficiency while recording, and 33 percent improvement in latency.

HUAWEI’S CHIP DESIGNS PUT AN EMPHASIS ON ACCELERATING AI PROCESSING
Huawei has doubled down on its AI processing aspirations, adding a dual NPU (neural processor unit) to the Kirin 980, which performs AI-assisted image recognition tasks at a rate of 4,500 images per minute. By the same measure, the Snapdragon 845 reaches 2,371 and Apple’s A11, which enjoys performance leads in other categories, gets only 1,458. AI also aids the Kirin 980’s power efficiency, as Huawei says it’s using it to more accurately and intelligently predict load requirements, making it more responsive to the power needs of the user — both when the chip needs to power up more cores and when it’s done its task and can save energy by slowing down.

The architecture of the Kirin 980 has eight cores: two are for so-called turbo performance, two are for long-term performance, and the last and smallest four are used to maximize power efficiency when not much is going on.

The Kirin 980 will offer the world’s fastest smartphone Wi-Fi speed, clocking in at 1,732Mbps, which is substantially higher than the Snapdragon 845’s best of 866Mbps with a Qualcomm modem or 1083Mbps with a third-party modem. So, by all metrics that matter to an end user, the new Kirin chip is shaping up to be a winner.

Huawei, with the Mate 20, and its sub-brand Honor, with the Magic 2, are going to be putting the Kirin 980 into retail devices by the end of the year.


Huawei promises its Kirin 980 processor will destroy the Snapdragon 845 - The Verge

 

[cirr]

AMOLED芯片国产化里程碑！吉迪思携手中芯北方发布首款QHD柔性智能机面板驱动芯片

2018-9-5

近日，深圳吉迪思携手晶圆代工企业中芯北方集成电路制造(北京)有限公司（以下简称:中芯北方）共同宣布，AMOLED智能手机面板驱动芯片正式进入量产，这将是真正意义上的大陆首款自主研发的AMOLED驱动芯片。芯片设计环节由国内先进显示驱动芯片设计领导者深圳吉迪思完成,芯片制造商为中国内地集成电路制造企业中芯北方，服务客户为国内显示面板界翘楚京东方,并于成都B7柔性线生产。

随着苹果iPhone也加入AMOLED的阵营，华为、小米也在OPPO、vivo之后加大了AMOLED机型的开发力度。根据CINNO Research手机产业链调研数据预测，2018年搭载柔性AMOLED屏的智能手机出货量有望翻倍增长至1.8亿部。AMOLED作为取代LCD的下一代显示技术，以其轻薄、柔性、更高的对比度等优势赢得了各大手机厂商的青睐。然而，产能规模的迅速扩张与国内OLED驱动芯片自主研发能力储备不足之间存在不平衡，AMOLED驱动芯片此前全部依赖进口，直到深圳吉迪思携手中芯北方正式量产该产品，首次打破大陆柔性AMOLED“缺芯少屏”的僵局。

CINNO Research副总经理杨文得分析，目前手机AMOLED面板的应用上虽然依旧是三星领先，但在中国面板厂逐渐在柔性AMOLED产品投注资源，我们认为采用AMOLED面板的智能型手机比重在今年将超过30%，而柔性(Flexible)的AMOLED面板出货将首次超过刚性(Rigid)。在这样特殊的产业结构下，智能型手机用的AMOLED面板驱动IC有绝对市场占有率的为三星自己，其它非三星阵营的AMOLED驱动IC还是在奋斗站稳一席之地的阶段，但随着产业供应链国产化的长程目标建立下，中国AMOLED面板驱动IC的成长将指日可待。

吉迪思此次量产的产品采用40纳米高压驱动工艺,首次采用32纳米SRAM单元，面积缩小20%，与同类驱动芯片相比，速度提高20%；可同时支持HD，FHD和QHD等多种显示模式，全面兼容三星、高通、苹果等多种智能手机应用处理器。

深圳吉迪思电子科技创始人田琪表示：“作为中国第一个于2016年5月在智能手机终端客户量产刚性HD AMOLED驱动芯片的公司，吉迪思早在2015年3月就正式开始和中芯合作。去年12月22日，采用32纳米SRAM的高压AMOLED工艺在中芯北方立项。在中芯北方的密切配合和全力支持下，仅用了8个月就达成了最先进的QHD柔性屏DDIC的量产，远远超出了吉迪思的预期。我们相信，AMOLED作为全球一线智能手机OEM力挺的下一代面板显示技术，在2019年会彻底爆发，出现S曲线最陡的一段，并在2020年超越LCD智能手机出货量。抓住该契机的唯一办法就是加速产品迭代，迅速打入一线品牌。我们期待和中芯北方即将开始的再次合作会更精彩。”

中芯北方总经理张昕表示: “40纳米显示技术平台集成了超低功耗、高压驱动工艺，适用于驱动AMOLED屏幕的芯片产品，能够很好地满足客户对功耗、面积和性能的要求。此次与吉迪思深度合作，共同推进AMOLED驱动芯片在中国的自主设计、制造以及商业化进程，意义重大。”

随着AMOLED技术不断走向成熟，应用范围愈加广泛，AMOLED显示产业出现急剧增长的形势。国际厂商纷纷逐鹿AMOLED市场,国内市场也力争抓住该领域先机。相信,在AMOLED市场需求量猛增的浪潮下,国产芯片将以此为契机、迎头赶上,争做AMOLED新领域的弄潮儿。

 

[cirr]

China 12-inch fab capacity set to boom

Claire Sung, Taipei; Jessie Shen, DIGITIMES

[Tuesday 14 March 2017]

Startup Huaian Imaging Device Manufacturer (HIDM) plans to build a 12-inch fab in Huaian, Jiangsu province, with production capacity set at 20,000 wafers monthly. Founded in 2016, HIDM specializes in the design and manufacture of CMOS image sensor devices.

http://www.digitimes.com/news/a20170311PD200.html

@Bussard Ramjet India?

德淮半导体（HiDM）进军高端智能手机市场 HR1630

2018-09-05 11:09:00 作者：佚名 出处 : 博客

近来, 国内中高阶手机的后置摄像头开始流行16MP配置, 今年发布的众多手机前摄也导入了16MP规格。预计2019年，16MP在前后摄的普及将成为市场趋势。就关键器件成熟度而言, 16MP无论是分辨率、低照度拍摄效果和PDAF性能都已超越13MP, 为用户升级提供了充足的底气。

德淮半导体(HiDM)今年推出16MP的HR1630采用了更小像素(1.0um)的堆栈式工艺, 其整体影像效果不但优于目前主流的1.12uM 13MP产品，甚至在暗态环境下的信噪比(SNR)也有更优异的表现 (如下图是在25Lux的测试结果)。

　　 HR1630(1.0uM) 主流13MP (1.12uM)

　　 25Lux / SNR:27dB 25Lux / SNR:24dB

HR1630可以轻松实现8mm(L) x 8mm(W) Auto Focus和 7.5mm(L) x 7.4mm(W) Fix Focus的模组尺寸, 更便于塑造美观的手机造型。

更令人惊艳的是, HR1630升级了第二代“微光快拍”PDAF技术，更高密度的PD分布和感光量，即使在低于20 lux的微光环境下, 相位对焦仍然能够快速稳定地工作, 大幅提升手机的夜间抓拍能力。此外，HR1630导入了专业级的DCG(Dual Conversion Gain)技术，在暗态下使用High CG提升信噪比(SNR)、高亮度下使用Low CG可提升动态范围(Dynamic Range)确保了全天候复杂场景的拍摄效果。

2018年，中国手机产业仍然受制于严重依赖海外供货的窘境，多次发生大范围缺货。德淮半导体(HiDM)已整装待发，将于不久的将来连续推出全系列中高端CIS产品，依托其强大的研发能力和全自主12英寸晶圆产能与工艺的保障，彻底改变中国中高端CIS产业链，确保中国手机产业始终立于世界领先地位。(校对/叨叨)

总投资450亿元的淮安德淮半导体有限公司，按照分批建设计划，目前已完成一厂项目建设，二厂项目已开工。一厂项目预计在今年10月份正式投产，可年产24万片12英吋集成电路芯片。

 

[cirr]

上海复旦微亿门级FPGA

复旦微最新推出的亿门级FPGA：7K325T，对标XIlinx公司的同型号产品XC7K325T

上海复旦微电子集团股份有限公司在FPGA领域有着近二十年的研究和发展。公司前期研制出的自主知识产权千万门级FPGA产品，突破了在传统集成电路设计基础上的高可靠性设计，经过测试，其高可靠性能处于国际领先地位。该系列产品已成功应用于我国卫星导航、载人航天等重大工程项目中，解决了我国高可靠FPGA禁运的难题。目前，公司研制的新一代自主知识产权亿门级FPGA产品，采用了全新的亿门级FPGA创新架构，并集成了专用超高速串并转换模块、高灵活可配置模块、专用数字信号处理模块、高速内部存储模块、可配置时钟模块等适用亿门FPGA应用的模块电路，其各类指标均已达国际同类产品先进水平。该亿门级系列产品的成功研制填补了国内超大规模亿门级FPGA的空白，可满足我国对国防、航空、航天、通信、医疗等领域FPGA器件的迫切需求。

 

[qwerrty]

德淮半导体（HiDM）进军高端智能手机市场 HR1630

2018-09-05 11:09:00 作者：佚名 出处 : 博客

近来, 国内中高阶手机的后置摄像头开始流行16MP配置, 今年发布的众多手机前摄也导入了16MP规格。预计2019年，16MP在前后摄的普及将成为市场趋势。就关键器件成熟度而言, 16MP无论是分辨率、低照度拍摄效果和PDAF性能都已超越13MP, 为用户升级提供了充足的底气。

德淮半导体(HiDM)今年推出16MP的HR1630采用了更小像素(1.0um)的堆栈式工艺, 其整体影像效果不但优于目前主流的1.12uM 13MP产品，甚至在暗态环境下的信噪比(SNR)也有更优异的表现 (如下图是在25Lux的测试结果)。

　　 HR1630(1.0uM) 主流13MP (1.12uM)

　　 25Lux / SNR:27dB 25Lux / SNR:24dB

HR1630可以轻松实现8mm(L) x 8mm(W) Auto Focus和 7.5mm(L) x 7.4mm(W) Fix Focus的模组尺寸, 更便于塑造美观的手机造型。

更令人惊艳的是, HR1630升级了第二代“微光快拍”PDAF技术，更高密度的PD分布和感光量，即使在低于20 lux的微光环境下, 相位对焦仍然能够快速稳定地工作, 大幅提升手机的夜间抓拍能力。此外，HR1630导入了专业级的DCG(Dual Conversion Gain)技术，在暗态下使用High CG提升信噪比(SNR)、高亮度下使用Low CG可提升动态范围(Dynamic Range)确保了全天候复杂场景的拍摄效果。

2018年，中国手机产业仍然受制于严重依赖海外供货的窘境，多次发生大范围缺货。德淮半导体(HiDM)已整装待发，将于不久的将来连续推出全系列中高端CIS产品，依托其强大的研发能力和全自主12英寸晶圆产能与工艺的保障，彻底改变中国中高端CIS产业链，确保中国手机产业始终立于世界领先地位。(校对/叨叨)

总投资450亿元的淮安德淮半导体有限公司，按照分批建设计划，目前已完成一厂项目建设，二厂项目已开工。一厂项目预计在今年10月份正式投产，可年产24万片12英吋集成电路芯片。

very nice. chinese CIS companies are either targeting at the very very high-end like gpixel or in mid-low end market like gallaxycore, smartsens, superpix, ruixin..

 

[Bussard Ramjet]

very nice. chinese CIS companies are either targeting at the very very high-end like gpixel or in mid-low end market like gallaxycore, smartsens, superpix, ruixin..

And neither are succeeding. The last I checked the combined market share of these companies doesn't enter even double digits.

Also, in imaging, the number of pixels is the most hyped, and useless feature to talk about. Sensor and Pixel Size, sensitivity, ISO, are all more important.

上海复旦微亿门级FPGA

复旦微最新推出的亿门级FPGA：7K325T，对标XIlinx公司的同型号产品XC7K325T

上海复旦微电子集团股份有限公司在FPGA领域有着近二十年的研究和发展。公司前期研制出的自主知识产权千万门级FPGA产品，突破了在传统集成电路设计基础上的高可靠性设计，经过测试，其高可靠性能处于国际领先地位。该系列产品已成功应用于我国卫星导航、载人航天等重大工程项目中，解决了我国高可靠FPGA禁运的难题。目前，公司研制的新一代自主知识产权亿门级FPGA产品，采用了全新的亿门级FPGA创新架构，并集成了专用超高速串并转换模块、高灵活可配置模块、专用数字信号处理模块、高速内部存储模块、可配置时钟模块等适用亿门FPGA应用的模块电路，其各类指标均已达国际同类产品先进水平。该亿门级系列产品的成功研制填补了国内超大规模亿门级FPGA的空白，可满足我国对国防、航空、航天、通信、医疗等领域FPGA器件的迫切需求。

Again, we will talk about this if it actually gains some market share.

I recently read an article where a Chinese industry expert was quoted as saying that China is pretty much non existent in commercial FPGA and GPUs.

德淮半导体（HiDM）进军高端智能手机市场 HR1630

2018-09-05 11:09:00 作者：佚名 出处 : 博客

近来, 国内中高阶手机的后置摄像头开始流行16MP配置, 今年发布的众多手机前摄也导入了16MP规格。预计2019年，16MP在前后摄的普及将成为市场趋势。就关键器件成熟度而言, 16MP无论是分辨率、低照度拍摄效果和PDAF性能都已超越13MP, 为用户升级提供了充足的底气。

德淮半导体(HiDM)今年推出16MP的HR1630采用了更小像素(1.0um)的堆栈式工艺, 其整体影像效果不但优于目前主流的1.12uM 13MP产品，甚至在暗态环境下的信噪比(SNR)也有更优异的表现 (如下图是在25Lux的测试结果)。

　　 HR1630(1.0uM) 主流13MP (1.12uM)

　　 25Lux / SNR:27dB 25Lux / SNR:24dB

HR1630可以轻松实现8mm(L) x 8mm(W) Auto Focus和 7.5mm(L) x 7.4mm(W) Fix Focus的模组尺寸, 更便于塑造美观的手机造型。

更令人惊艳的是, HR1630升级了第二代“微光快拍”PDAF技术，更高密度的PD分布和感光量，即使在低于20 lux的微光环境下, 相位对焦仍然能够快速稳定地工作, 大幅提升手机的夜间抓拍能力。此外，HR1630导入了专业级的DCG(Dual Conversion Gain)技术，在暗态下使用High CG提升信噪比(SNR)、高亮度下使用Low CG可提升动态范围(Dynamic Range)确保了全天候复杂场景的拍摄效果。

2018年，中国手机产业仍然受制于严重依赖海外供货的窘境，多次发生大范围缺货。德淮半导体(HiDM)已整装待发，将于不久的将来连续推出全系列中高端CIS产品，依托其强大的研发能力和全自主12英寸晶圆产能与工艺的保障，彻底改变中国中高端CIS产业链，确保中国手机产业始终立于世界领先地位。(校对/叨叨)

总投资450亿元的淮安德淮半导体有限公司，按照分批建设计划，目前已完成一厂项目建设，二厂项目已开工。一厂项目预计在今年10月份正式投产，可年产24万片12英吋集成电路芯片。

Such articles only fool people who know nothing about the field. Pixel Number is one of the least important parameters in actual image quality.

If pixel number were so important, Chinese company Galaxy core would already be beating Sony.

It is in fact detrimental to you yourself if you overestimate your own technological progress. There is a huge and very long distance to go, which might be covered, who knows, but is definitely not covered as of now.

100G silicon photonics chip for optical transceiver put into production developed by accelink and government research institutes.


-------------------------------------------------------------------------
中国自主研发的首款商用100G硅光芯片正式投产
来源: 人民网-人民日报 作者: 范昊天 2018-08-30 14:04

[导读]记者从中国信息通信科技集团获悉：日前，我国自主研发的首款商用“100G硅光收发芯片”正式投产使用。

原标题：我国首款商用100G硅光芯片投产

本报武汉8月29日电 （记者范昊天）记者从中国信息通信科技集团获悉：日前，我国自主研发的首款商用“100G硅光收发芯片”正式投产使用。该系列产品支持100—200Gb/s高速光信号传输，具备超小型、高性能、低成本、通用化等优点，可广泛应用于传输网和数据中心光传输设备。

据介绍，该款商用化硅光芯片由国家信息光电子创新中心、光迅科技公司等单位联合研制。在一个不到30平方毫米的硅芯片上，集成了包括光发送、调制、接收等近60个有源和无源光元件，是目前世界上集成度最高的商用硅光子集成芯片之一。

国家信息光电子创新中心专家委员会主任余少华说，此次100G硅光芯片的产业化商用，是硅光技术领域的新突破，表明我国已经具备了硅光产品商用化设计的条件和基础。借助集成电路已大规模商用的CMOS工艺平台实现硅光芯片的生产制造，可以有效解决我国高端光电子芯片制造能力薄弱、工艺能力不足的问题
http://finance.zijing.org/2018/0830/763436.shtml

------------------
huawei's caliopa

http://www.caliopa.com/technology

Just so that one knows here, 100G and 400G products are already in commercial production right now, and next generation is in development.

So Accelink is two generations late to the product. (Which also pretty much defines the whole existence of Accelink)

 

[sweetgrape]

And neither are succeeding. The last I checked the combined market share of these companies doesn't enter even double digits.

Also, in imaging, the number of pixels is the most hyped, and useless feature to talk about. Sensor and Pixel Size, sensitivity, ISO, are all more important.



Again, we will talk about this if it actually gains some market share.

I recently read an article where a Chinese industry expert was quoted as saying that China is pretty much non existent in commercial FPGA and GPUs.




Such articles only fool people who know nothing about the field. Pixel Number is one of the least important parameters in actual image quality.

If pixel number were so important, Chinese company Galaxy core would already be beating Sony.

It is in fact detrimental to you yourself if you overestimate your own technological progress. There is a huge and very long distance to go, which might be covered, who knows, but is definitely not covered as of now.




Just so that one knows here, 100G and 400G products are already in commercial production right now, and next generation is in development.

So Accelink is two generations late to the product. (Which also pretty much defines the whole existence of Accelink)

You just should know that, More Chinese companies are investing huge money in IT industry, we are on the way, It is very obvious, compared with best one, there are a huge gap, but at least, we start and speed up, that's enough.

There are only two countries in the world: China and a foreign country(Including India).